SynOmega Skill

August 20, 2026 · View on GitHub

An agent Skill for SynOmega — the retrosynthesis and reaction-prediction toolkit on PyPI (docs). It teaches Claude Code, OpenClaw and other coding agents to use the synomega Python package across its six capabilities: single-step retrosynthesis (product → reactants), single-step forward prediction (reactants → product), multi-step route planning, a continuous synthesizability score (SynScore), reaction-plausibility screening, and multi-component evolution (growing a forward synthesis network from a set of reactants).

The skill runs synomega locally — pip install synomega plus a trained model and a building-block file. It does not depend on any hosted service.

Install

OpenClaw / ClawHub

clawhub install synomega

Claude Code (manual)

mkdir -p ~/.claude/skills/synomega
curl -fsSL https://raw.githubusercontent.com/zbc0315/synomega-skill/main/SKILL.md \
  -o ~/.claude/skills/synomega/SKILL.md
curl -fsSL https://raw.githubusercontent.com/zbc0315/synomega-skill/main/scripts/synomega_run.py \
  -o ~/.claude/skills/synomega/synomega_run.py

Prerequisites

pip install "synomega[gnn]"        # the package (neural backend)

That's it — it works out of the box. The default pretrained model and building-block stock download automatically on first use (into ~/.cache/synomega); run synomega download to pre-fetch them. Downloads come from the nearest mirror (USTC GitLab in China, or GitHub), auto-selected by latency. To use your own checkpoint/stock instead, set SYNOMEGA_MODEL / SYNOMEGA_STOCK.

Use

Ask your agent things like:

  • "Can paracetamol be synthesized? How hard?"
  • "Propose a synthesis route for CC(=O)Nc1ccccc1O."
  • "What reactants could give this molecule in one step?"
  • "What product do acetic acid and benzylamine give?"
  • "Evolve a forward network from acetophenone + formaldehyde + dimethylamine."

Or call the bundled helper directly (one JSON-printing command per capability):

python scripts/synomega_run.py single-step  "CC(=O)Nc1ccccc1O" --top-k 10     # product -> reactants
python scripts/synomega_run.py forward      "CC(=O)O.NCc1ccccc1" --top-k 5    # reactants -> product
python scripts/synomega_run.py plan         "CC(=O)Nc1ccccc1O" --max-depth 5  # multi-step route
python scripts/synomega_run.py score        "CC(=O)Nc1ccccc1O" --max-steps 5  # synthesizability (SynScore)
python scripts/synomega_run.py evolve       "CC(=O)c1ccccc1.C=O.CNC" --max-depth 3 --score-threshold 0.01

plan and score take --exclude-target (treat the target as not purchasable even if it is a catalogue molecule, so it is not trivially "solved" in zero steps). Reaction-plausibility screening is an env toggle: SYNOMEGA_PLAUSIBILITY=1. See SKILL.md for the full option list and output shapes.

Contents

FilePurpose
SKILL.mdthe skill definition (frontmatter + instructions)
scripts/synomega_run.pyloads model + stock from env vars, runs any capability (single-step / forward / plan / score / evolve), prints JSON

License

MIT — see LICENSE.